Classification and MT

Two new MT plug-ins have come into being since I posted about the way I use the category field here. They offer two different ways of classifying posts. MT-categoryfaceted by Timothy Appnel is a bottom-up method and MT-subcategories by David Raynes is hierarchical and a top-down method. Take your pick.

Hierarchical classification is like file systems on our computers, Yahoo!, and the Dewy Decimal System. We often consider ideas in terms of topics and their subtopics so it's pretty natural to think of sorting things out this way. However, you usually run into difficulties because hierarchies are inflexible and don't deal with evolution well.

Faceted classification is bottom-up. I've never had much of a problem grasping it since I spent several years studying art history and archaeology. In that field, you start building your body of knowledge in a rather bottom-up manner. You spend a lot of time identifying objects as such:

title: Un dimanche après-midi à l'Ile de la Grande Jatte
date: 1884-86
artist: Georges Seurat, French 1859-1891
medium: Oil on canvas
size: 207.5 x 308 cm
movement: post-impressionism
provenance: The Art Institute of Chicago

If you wanted to read about it, you could look in a book about Seurat, post-impressionism, oil painting, The Art Institute of Chicago, or sofa-sized art. You can search for it from within whichever facet you want to consider.

Christina in Information Architecture: Blueprints for the Web mentions that using faceted classification works best if what you are classifying is fairly homogenous. I like the warning she gives after discussing the faceted classification used on epicurious. It mirrors what I've observed.

Even homogenous content can have pitfalls. One such is helping users who want to be sure they've got a comprehensive view; for example, suppose they want to see every recipe for tomato soup. They may become anxious because it will be hard to know when they've seen them all-- especially when they see the same recipe cropping up in a few categories. For example, they might see gazpacho in cold soups and summer meals. Searchers may wonder if they are retreading the same ground.

There's been a lot of conversation about using faceted classification for information retrieval on the web in the last two years. I have links to some of it under classification. Here are two you might want to look at:

• After the Dot-Bomb: Getting Web Information Retrieval Right This Time"
  Discussion about it on SIGIA-L
• Ramana Rao's Information Flow Issue #4 August 2002

Something Victor mentioned is that using a robust and appropriate classification system is one thing, but how you then present it so that users can leverage it is another. Keep in mind that all of this arranging of stuff is so you and your users can find it.

I've never been able to fix the nav here so that I could do things like "see all articles about IA from January 2003." However, with a little help from Peter I did make a template that gives me an xfml output. It's interesting to upload that output into FacetMap and browse it.